The population of the study area is genetically quite uniform and no different ethnic groups are living there. Screening for T2DM takes place in a totally different environment in high income countries compared to countries in the category of low or middle income, such as Thailand. So-called primary care in high income countries usually is a ‘passive’ service provided by the ambulatory care section of the health delivery system, in that individuals are encouraged to see their general practitioners in their clinic and they only undergo a test according to the decision of the particular clinician they consult. In Thailand the MoPH has initiated a more active screening effort in rural areas, where the general population is being approached by VHVs to assemble at a central spot within their community after a nights fasting to be screened by the staff of the PCU in charge of the village using glucometer kits. Whatever the MoPH advocates has to be easy to perform, must be time and cost effective, and care has to be taken not to overload the clinical arm of the health delivery system, such as district and provincial hospitals, with a heavy load of false positive cases indicated by a low PPV.
It is not feasible for the understaffed and overworked health officials at the sub-district level to take more than one CBG sample and come back to the same village after about one week and take another sample. Also supervision of the staff by their superiors from the provincial level is not possible when doubling the work. Also, to actually relate the result from the first and the second sample to one and the same individual needs accuracy which cannot be always expected from a public health nurse with a two years training let alone from a VHV. Similarly, to conduct an oral glucose tolerance test (OGTT) on a population basis is beyond feasibility. One round in one district might involve testing more than 7,000 individuals in 119 villages.
This study attempted to follow the screening procedures of the Thai health delivery service of the participating district as much as possible in order to evaluate its achievements but also its constraints. According to the study plan it was not intended to draw a representative sample of adults from the district. Still, the proportion of individuals found to have fasting blood glucose concentrations of 7 mmol/L (126 mg/dl) and above was similar to the T2DM prevalence estimated in nationwide surveys, which is reported to be 9.6%, with 4.8% newly diagnosed cases in 2000 [9
] and 6.7% in 2004 [10
]. The latter estimate was age standardized based on the Thai national population. In 1995 a representative sample of three districts of a neighboring province (Khon Kaen) of the study area detected a prevalence of T2DM of 11.9% with a 95% C.I. of 8.4 to 16.5 by using the OGTT [11
]. This figure compares well to the proportion of individuals found in this study to have HbA1c levels of 7% and over, which is 10.4%.
Since this study is restricted to reporting the validity of screening tests to detect T2DM diseased patients, the proportion of Impaired Fasting Glucose (IFG), in the range of 6.1 mmol/L (101 mg/dl) to 6.9 mmol/L (124 mg/dl), has not been reported. As outlined in the methodology section, in the real field situation individuals found to have fasting blood glucose levels of 6.1 mmol/L (101 mg/dl) to 6.9 mmol/L (124 mg/dl) are advised by the staff to come to the nearest PCU in approximately 6 months to be rechecked, at which time it may be assumed that they will have a fasting blood sugar level over 7 mmol/L (126 mg/dl) if they are indeed early T2DM cases. Since it is estimated that half of all individuals having T2DM in Thailand remain undetected [9
], the aim of the health authorities at this time is to concentrate clinical attention on those being found to already have T2DM. Using fasting blood glucose with a cutoff point of 7.0 mmol/L (126 mg/dl) as a valid test for the diagnosis of T2DM is recommended by the World Health Organization (WHO) [12
]. Using CBG as a tool for screening and VPG determinations as a final diagnosis results in a sensitivity of over 80%, a specificity of over 95% and a PPV of 70%. Using CBG as a screening indicator against VPG as standard, the results as far as sensitivity, specificity and PPV are concerned seems to be acceptable. According to the ROC curve of this investigation the optimal cutoff point would be 5.6 mmol/L (101 mg/dl), which is in accordance with the findings of Nitiyanant et al. [13
], who recommended a cutoff point between 5.6 mmol/L (101 mg/dl) and 6.0 mmol/L (108 mg/dl) for the Thai population.
When CBG values are used as a screening tool and VPG as a reference, less than 20% of those screened remained with undetected T2DM. However, when selecting HbA1c as the reference and CBG as the screening tool, 54.4% of suspected T2DM cases are not identified and the proportion of false negative results increases to 41.5%. In the case where VPG is the screening tool and HbA1c is the reference, 60.3% of T2DM cases are not identified and the curative center is unnecessarily burdened with 43.2% of the cases, which are false positives.
HbA1c was recommended as a diagnostic tool as far back as 1976 by Koenig et al., [14
] and its use has increased over time in clinical settings in high income countries. For example a survey undertaken in Ontario (Canada) found that in 2005 HbA1c were determined in approximately 500,000 individuals without diabetes and for more than 480,000 cases with diagnosed diabetes [16
]. In the same year, 37% of the Ontario adults without T2DM were tested by serum blood glucose and less than 1% of clinicians used the OGTT. The controversial discussion about the application of HbA1c for either screening or clinical diagnosis intensified after the American Diabetes Association (ADA) and the European Association for the Study of Diabetes recommended the use of HbA1c to diagnose T2DM [17
]. That recommendation was included in the ADA’s suggestions [18
]. Recently also the IDF and WHO pointed towards HbA1c as diagnostic tool for T2DM [19
], recommending a cutoff point for the diagnosis of less then 6.5%. Despite the widespread use of HbA1c in clinical settings, different cutoff points were recommended for the diagnosis of T2DM, besides ≥6.5% from ADA and WHO [18
] also ≥7% from other sources, such as the recommendation of Davidson and Schriger [20
] based on reassessing 2,712 individuals ‘without diabetes history’ and the use of ≥7% by the New York City A1c Registry [21
]. The distribution of HbA1c in a sample of 323 healthy Thai individuals determined the 2.5 and the 97.5 percentiles to be 2.9 to 4.9% respectively, with a mean VPG of 5.11 mmol/L (92.1 mg/dl) [22
]. The ROC curve for HbA1c as a reference with a 7% cutoff point resulted in an optimal screening test of 5.03 mmol/L (90.1 mg/dl), which is similar to the above quoted findings. However, the associated specificity of 50% would be unacceptable for the clinical settings. Despite this the standard cutoff point set by the MoPH was a VPG of 7 mmol/L (126 mg/dl) and it was against the objective of the study to overrule this.
The somehow ‘low’ prevalence of chronic diseases usually results in a low PPV indicating the burden of the collaborating hospitals to check a high number of individuals with false positive results. In this situation a slight increase in prevalence results in a remarkable increase in PPV. That could have been achieved by selecting individuals with a high risk of having T2DM, such as elevating the cutoff point for age to 45 years and selecting only over nourished and obese individuals as well as those with a family history of T2DM for being eligible for screening. Again this was against the objective of the study since the MoPH only excluded individuals with an age of <35 years to be the target for screening. PPV also might increase due to lowering the cutoff point of the clinical reference, in this case from HbA1c values of ≥7% to ≥6.5% because the lower cutoff point will increase the prevalence and by this PPV. In fact a cutoff point of 6.5% HbA1c increases PPV for CBG to 67.9% and VPG to 70.4% (unpublished data).
Another important aspect however is to achieve an optimal sensitivity of the screening method. Using HbA1c ≥6.5% as cutoff point sensitivity for both screening indicators decreased, from 45.6% to 31.0% for CBG and from 39.7% to 28.4% for VPG (unpublished data). It might be argued that the low sensitivity by selecting 6.5% HbA1c as clinical reference is due to a high proportion of screened individuals actually not having the disease and consequently normal CBG and VPG values. The intention was to increase the probability that the clinical reference, here an HbA1c of 7% and above really identify a diseased person.
According to the results of this study, the determination by CBG is better than determination of VPG, however, the fact remains that the associated sensitivity of 45% for CBG results is rather low, and it is even lower for the VPG determinations.
Weighing the pros and cons of the use of HbA1c in so-called ‘developing countries
’, the authors of a recent publication concluded that HbA1c should not be used for the diagnosis of T2DM in these settings [23
]. The authors argued that the threshold for diagnosis, the cost of the test, the absence of a standardization network and the low sensitivity were all factors that make HbA1c an inappropriate tool for T2DM diagnosis. As far Thailand is concerned, HbA1c is often used in hospitals in monitoring T2DM in patients. Presently district and provincial hospitals have equipment and financial means to determine HbA1c correctly and each has a responsible laboratory or will be included into a standardization network. Therefore, a low sensitivity of CBG results in itself might not be a sufficient argument against the use of HbA1c as clinical reference in Thailand. Even in rural Thailand, villagers are well aware of at least the essentials of T2DM and know that blood glucose levels correlate to dietary intake [24
An additional argument against the indiscriminate use of HbA1c is that the variation of the glycated hemoglobin also depends on HbA1c variants and ethnicity [25
]. Other factors associated with plasma glucose levels have been reviewed recently by Herman [29
] and Gomez-Perez et al. [23
] and include anemia and iron deficiency as well as abnormal hemoglobins. Testing the applicability of HbA1c for Thai individuals as reported above [22
] did not suggest that the Thai population was different from Caucasians as far as HbA1c levels are concerned. However, the individuals tested had been derived from a population in Bangkok and not from a population of the Northeast of the country, where anemia and iron deficiency are common and Thalassemia is a public health problem [30
]. Haemoglobin E (Hb E) is the most common trait for hemoglobinopathies prevalent in the Northeast of Thailand [32
]. Adjusted HbA1c values were higher in a group of Asians in comparison to white individuals with Impaired Glucose Tolerance (IGT) but the adjusted HbA1c difference was insignificant (5.78% versus 6% respectively) [29
]. Erythrocytes are vulnerable in the Thalassemia disease [33
] and the decrease of the mean erythrocyte age falsely lowers HbA1c test results [34
]. Weak but significant differences between HbA1c levels of ‘normal’ subjects and hetero- as well as homo-zygote Hb E carriers have been determined previously [35
]. However, the median values of HbA1c levels of the Hb E groups have been below the so called ‘normal’ controls. Therefore, the high number of false negative results when using HbA1c as clinical reference and fasting plasma glucose levels as screening test should not be due to the prevalence of thalassemia and Hb E.
An additional hint that false negative screening results against HbA1c might not be related to haematological issues but might be an indication of the VPG inadequacy as a screening tool is the result of a study undertaken some years ago, close to the area where this investigation has been based [11
]. At that time (1995), HbA1c was not used as a reference, rather OGTT was preferred, with the cutoff point of 11.1 mmol/L (200 mg/dl), and the screening tool was fasting blood glucose, with a cutoff point of 7.8 mmol/L (140 mg/dl). These settings resulted in a sensitivity of 43.7%, specificity 89.8% and the PPV 37.8%. It seems that in the case of Thailand the use of both HbA1c and OGTT as final diagnostic tools results in similarly low sensitivities, as long as fasting blood glucose measurement is applied as screening tool.
Chronic kidney diseases may also affect HbA1c levels through increased blood urea nitrogen levels, causing the formation of carbamylated haemoglobin which cannot be differentiated in the determination from HbA1c [36
]. In a recent publication from Thailand the proportion of T2DM patients with microalbuminuria was found to be about 40% [37
]. Therefore, it cannot be excluded that some of the cases of undetected T2DM within the population screened for this study may have microalbuminuria. However, it is unlikely that undetected chronic kidney disease already developed to such an extent that it could have influenced the level of HbA1c.
It has been reported recently that HbA1c is insufficient to detect T2DM in the case of early diabetic status [38
]. The present situation in Thailand mainly demands the identification of those who have been suffering from T2DM for quite some time already, so that this limitation of HbA1c at the present time may be overlooked.
Further investigations are necessary to recognize the reasons for the low sensitivity of the screening indicators. It has been reported recently that a high proportion of T2DM patients obviously do not feel comfortable to admit that they do not follow the advice of health personnel [39
]. In general a Thai villager is quite well informed about the basics of diabetes mellitus [24
]. It might well be that, in an ill attempt to have a favorable screening result, participants might limit energy and carbohydrate intake for a number of days before screening, since the event is arranged about a week in advance. The failure to identify individuals with yet undetected T2DM might be less a technical problem related to HbA1c, CBG or VGP considering that a similar investigation using OGTT as reference standard some years before resulted in a similarly insufficient VPG sensitivity [11
]. The underlying reason for the observed discrepancies between VPG and a clinical standard, either being OGTT or HbA1c, might be due to the fact that the villagers can somehow influence the VPG, in reducing carbohydrate and calorie intake before screening, but not the latter tests. What the villagers may not be aware of is that HbA1c values reflect the status of glucose levels over approximately a three months period. In the context of Thai culture, villagers may be sensitive about being confronted by the health staff after being checked and found with unfavorable results. Health promotion experts might further investigate this phenomenon and try to change the attitude of the villagers.